Platform-specific tracking is necessary because answer behavior, source presentation, and collection availability differ across products.
Define the Claude panel
Include category, product discovery, comparison, implementation, trust, and audience-specific prompts. Keep a fixed core panel for trends and a separate candidate library for new questions.
Preserve answer-level facts
Record:
- successful, failed, and excluded response states;
- exact brand and alias matches;
- recommendation position and TOP3 status;
- Positive, Neutral, or Negative Sentiment;
- factual description issues;
- competitors mentioned in the same answer;
- links or source evidence where available.
Do not infer missing citations when the interface or collection route does not expose them.
Compare Claude with other platforms carefully
Use the same business questions where possible, but keep platform results separate. A difference may reflect answer behavior, source availability, collection method, or timing. It is not automatically proof that one platform “prefers” a brand.
| Comparison field | Requirement |
|---|---|
| Prompt | Same wording or a documented platform-specific adaptation |
| Market and language | Same target conditions |
| Date window | Comparable observation period |
| Success status | Platform failure excluded from visibility denominators |
| Entity rule | Same brand aliases and competitor definitions |
| Sentiment | Same Positive, Neutral, or Negative review standard |
| Sources | Report platform-specific availability rather than filling gaps |
Act on stable gaps
Repeated absence across important prompts can justify a page or positioning review. Inaccurate descriptions require entity and source correction. Competitor citations can reveal stronger official owners or external evidence. High volatility calls for more samples before implementation.
Brand absent, competitors present
Verify collection success and entity matching, then inspect which competitor roles and evidence appear. Create a new owner only if the question represents a distinct, valuable task.
Brand present, description inaccurate
Separate tone from factual accuracy. Identify the exact unsupported statement and reconcile official pages, documentation, profiles, and high-value external sources.
Source evidence unavailable
Keep the answer observation, but mark Citation provenance unavailable. Do not manufacture a source gap from a platform field the collection route did not expose.
Retest without changing the question
Save the page or source version, publication date, and intended prompts. Repeat the same Claude panel under comparable conditions. If wording changes are required, create a new prompt version rather than overwriting history.
UnderAI availability
UnderAI GEO Workspace’s current international default set includes ChatGPT, Google AI Overviews, Google AI Mode, Perplexity, and Gemini. Claude availability is not stated as a default. Confirm whether Claude can be enabled for your project and market before procurement.
If Claude collection is unavailable, UnderAI can still apply the same audit method to supplied, lawfully collected answer evidence without representing missing platform cells as zero.
Ask about platform coverage or review AI Search Monitoring.
Current UnderAI evidence boundary
Claude was not included in UnderAI’s July 2026 fixed Panel. The baseline contains ChatGPT, Perplexity, Gemini, and failed Google AI Mode and AI Overview routes; it cannot support a Claude visibility percentage.
That absence is part of the methodology, not a reason to fill the page with an inferred score. A Claude project should establish a separate baseline only when the collection route, market, model or product surface, source fields, and repeated Prompt set have been verified.
Claude-specific evidence record
When Claude web search is used, preserve:
- exact prompt and market conditions;
- whether web search ran;
- search success, empty result, or tool error;
- final response text;
- every source URL and title;
- cited text or source location when the route exposes it;
- brand role, Sentiment, and factual accuracy;
- repeated-run stability.
Do not compare a Claude API web-search run directly with a consumer product answer or another platform unless the route difference is disclosed.
Anthropic source documentation
Anthropic documents that Claude’s web search tool can access current web content and returns cited sources; it also exposes distinct error states inside the response body. See the Claude web search tool documentation. UnderAI uses those route-specific states when scoping a Claude measurement project rather than promising universal coverage.
Frequently asked questions
Does UnderAI currently guarantee Claude tracking?
No. Claude is not part of the documented current international default set. Confirm project availability before purchase.
Can I compare Claude and ChatGPT scores directly?
Only after aligning prompts, dates, markets, success rules, and metric formulas. Preserve platform-level results even then.
Does no source link mean Claude used no sources?
Not necessarily. Report what the collected interface exposes; do not infer unavailable provenance.
What if Claude tracking is not enabled in UnderAI?
Confirm platform availability before the project begins. If it is unavailable, results can remain outside the product panel or be analyzed from separately supplied evidence; they must not be represented as live UnderAI collection.